Development of novel computational protocols for the design of efficient enzymes

ENG- Enzymes are natural catalysts that speed up biochemical reactions efficiently, but their use in industry is often limited because they are usually specific to certain substances and reactions. Changing enzymes to work in new ways under certain conditions is difficult. One experimental method, Directed Evolution (DE), has shown remarkable results. However, it is expensive, time-consuming, and does not always explain why certain changes work better. Computational techniques present new possibilities for enzyme design. These methods help scientists understand how enzymes work and how they change their shape during reactions. For example, Molecular Dynamics (MD) simulations and methods of fast discovery of new shapes like accelerated MD or metadynamics allow researchers to explore a broader range of enzyme conformations and behaviors. However, these methods tend to start with a fixed enzyme structure, which can limit their ability to find new shapes. Statistical techniques like Principal Component Analysis (PCA) and Time-lagged Independent Component Analysis (TICA) help simplify complex data about enzyme movements but do not always show which parts of the enzyme are important for these changes. This thesis presents new computational strategies to better explore the range of enzyme structures and identify key sites for modification to create enzymes with new properties. Tools like the Shortest Path Map (SPM) web server, improvements in Deep Learning (DL) methods, and a new approach using Alphafold2 (AF2) for predicting enzyme shapes are herein discussed. It also showcases practical applications, such as modifying an enzyme called hydroxynitrile lyase (HNL) to become an efficient esterase (EST) through targeted mutations. These findings highlight the potential of computational methods to advance the design of new and improved enzymes